Using Gene-Set Analysis To Gain Biological Knowledge Based On GWAS Results

To date, many gene set analysis (GSA) approaches have been developed for identifying differentially expressed gene sets using microarray data. However, these methods are not directly applicable to RNA-Seq data due to intrinsic difference between two data structures. When testing the differential expression of gene sets, there is a critical assumption that the members in each gene set are sampled independently in most GSA methods. It means that the genes within a gene set don’t share a common biological function. The aim of this paper is twofold. First, we propose a powerful yet simple extension to GSA methods based on the de-correlation (DECO) algorithm that properly remove the correlation bias in the expression of each gene set. We then study the performance of our proposed method compared with other GSA methods through a real RNA-Seq dataset and simulation studies under various scenarios combining with four commonly used normalization methods. Second, we discuss the effect of the complex correlation structure of gene sets on four normalization methods. As a result, we found that our proposed method outperforms the others in terms of Type I error rate and empirical power. A comparative study on a public data showed that gene sets identified by our proposed method have better concordance with biological confirmed pathways than other methods.

Gene-set analysis (GSA) aims to identify sets of differentially expressed genes by a phenotype in DNA microarray studies. Challenges occur due to the salient characteristics of the data: (1) the number of genes is far larger than the number of observations; (2) gene expression measurements, especially within each gene set, can be highly correlated; and (3) the number of gene sets that can be examined is large and increasing rapidly. These challenges call for gene-set testing procedures that have both efficiency in computation for large GSAs and high power in the presence of the high correlation. We propose a new GSA approach called Linear Combination Test (LCT), incorporating the covariance matrix estimator of gene expression into the test statistic. The proposed LCT and two other GSA methods, a modification of Hotelling's T2 using a shrinkage covariance matrix and our SAM-GS (Dinu et. al. 2007), the two methods that have been reported by Tsai and Chen (2009) to perform best in terms of power, are evaluated in simulation studies and a real microarray study. The LCT method is more computationally efficient than the modified Hotelling's T2 and approximates the superb power of the modified Hotelling's T2. LCT is slightly faster than SAM-GS, but more powerful, due to incorporating the covariance matrix estimator. An extra step to enhance the interpretation of GSA results is also proposed in the form of a hierarchical LC (HLC) testing procedure, providing scientists useful hierarchical information on gene sets that LCT identified as differentially expressed.

Cancer initiation and progression are likely caused by the dysregulation of biological pathways. Gene set analysis (GSA) could improve the signal-to-noise ratio and identify potential biological insights on the gene set level. However, platforms exploring cancer multi-omics data using GSA methods are lacking. In this study, we upgraded our GSCALite to GSCA (gene set cancer analysis, http://bioinfo.life.hust.edu.cn/GSCA) for cancer GSA at genomic, pharmacogenomic and immunogenomic levels. In this improved GSCA, we integrated expression, mutation, drug sensitivity and clinical data from four public data sources for 33 cancer types. We introduced useful features to GSCA, including associations between immune infiltration with gene expression and genomic variations, and associations between gene set expression/mutation and clinical outcomes. GSCA has four main functional modules for cancer GSA to explore, analyze and visualize expression, genomic variations, tumor immune infiltration, drug sensitivity and their associations with clinical outcomes. We used case studies of three gene sets: (i) seven cell cycle genes, (ii) tumor suppressor genes of PI3K pathway and (iii) oncogenes of PI3K pathway to prove the advantage of GSCA over single gene analysis. We found novel associations of gene set expression and mutation with clinical outcomes in different cancer types on gene set level, while on single gene analysis level, they are not significant associations. In conclusion, GSCA is a user-friendly web server and a useful resource for conducting hypothesis tests by using GSA methods at genomic, pharmacogenomic and immunogenomic levels.

The focus of analyzing data from microarray experiments has shifted from the identification of associated individual genes to that of associated biological pathways or gene sets. In bioinformatics, a feature selection algorithm is usually used to cope with the high dimensionality of microarray data. In addition to those algorithms that use the biological information contained within a gene set as a priori to facilitate the process of feature selection, various gene set analysis methods can be applied directly or modified readily for the purpose of feature selection. Significance analysis of microarray to gene-set reduction analysis (SAM-GSR) algorithm, a novel direction of gene set analysis, is one of such methods. Here, we explore the feature selection property of SAM-GSR and provide a modification to better achieve the goal of feature selection. In a multiple sclerosis (MS) microarray data application, both SAM-GSR and our modification of SAM-GSR perform well. Our results show that SAM-GSR can carry out feature selection indeed, and modified SAM-GSR outperforms SAM-GSR. Given pathway information is far from completeness, a statistical method capable of constructing biologically meaningful gene networks is of interest. Consequently, both SAM-GSR algorithms will be continuously revaluated in our future work, and thus better characterized.

High throughput gene expression technologies have been widely used in many biological fields. Typical analysis of gene expression data is to find similarly expressed gene groups by clustering approaches and to identify differentially expressed genes by statistical approaches. The analysis, however, still has a difficulty in interpreting molecular level interaction or signaling transduction based on prior biological information. Recently, a Gene Set Analysis (GSA) approach was developed by a MIT group, which paved the first way for inferring molecular pathway mechanisms behind differentially express genes among sample groups. Current GSA approaches do not take hierarchical regulation among gene entries based on prior pathway information (e.g., KEGG pathways) into consideration. Our proposed approach is that GSA can be expanded not only to reflect the hierarchical structures among genes but also to identify specific subpathways that statistically agree with gene expression data as well as that could explain molecular level mechanism differences between two sample groups. We obtained the KEGG pathways (http://www.genome.jp/kegg/pathway.html) of which nodes and edges were taken into consideration by a probabilistic model. A statistic was calculated for each subpathway in every KEGG pathway based on the model. We identified significant subpathways in an expression dataset.

Model studies in mice indicate that the severity of alcohol withdrawal is associated with polymorphic variation and expression of the MPDZ gene. Current knowledge about variation in the human MPDZ gene is limited; however, our data indicate its potential association with alcohol dependence. The multi-PDZ protein is an important part of the N-methyl-D-aspartate (NMDA)-dependent α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid (AMPA) receptor trafficking cascade that controls glutamate-related excitatory neurotransmission. To investigate association of variation in the NMDA-dependent AMPA trafficking cascade with alcohol dependence, we performed a gene-set (pathway) analysis using single nucleotide polymorphism (SNP) data from the Study of Addiction: Genetic and Environment. Rather than testing for association with each SNP individually, which typically has low power to detect small effects of multiple SNPs, gene-set analysis applies a single statistical test to evaluate whether variation in a set of genes is associated with the phenotype of interest. Gene-set analysis of 988 SNPs in 13 genes in the pathway demonstrated a significant association with alcohol dependence, with P < 0.01 for the global effect of variation in this pathway. The statistically significant association of alcohol dependence with genetic variation in the NMDA-dependent AMPA receptor trafficking cascade indicates a need for further investigation of the role of this pathway in alcohol dependence.

To investigate the functional significance of genetic risk loci identified through genome-wide association studies (GWASs), genetic loci are linked to genes based on their capacity to account for variation in gene expression, resulting in expression quantitative trait loci (eQTL). Following this, gene set analyses are commonly used to gain insights into functionality. However, the efficacy of this approach is hampered by small effect sizes and the burden of multiple testing. We propose an alternative approach: instead of examining the cumulative associations of individual genes within a gene set, we consider the collective variation of the entire gene set. We introduce the concept of gene set QTL (gsQTL), and show it to be more adept at identifying links between genetic risk variants and specific gene sets. Notably, gsQTL experiences less susceptibility to inflation or deflation of significant enrichments compared with conventional methods. Furthermore, we demonstrate the broader applicability of shared variability within gene sets. This is evident in scenarios such as the coordinated regulation of genes by a transcription factor or coordinated differential expression.

Microarray, developed at the end of the last century, is one of the most significant achievements in biotechnology history. One of the most important applications of microarrays is to uncover the signature genes characterizing different samples or biological states. This application was initially used in individual gene analysis, but recently has been applied to gene set analysis. We compare these two approaches and compare the performance using JProGO (Java Tool for the Functional Analysis of Prokaryotic Microarray Data using the Gene Ontology). JProGO is able to examine microarray data from more than 20 different samples of prokaryotic and thus practical analysis of gene behavior data with regards to Gene Ontology (GO) is made possible. Two Escherichia coli datasets are used to compare the performance of the two analyses. The results show that the gene set analysis is more powerful because of its ability to detect the GO terms according to biological expectation. This research will provide a new platform and opportunities for further research or studies in microarray based gene sets.

Functional and genomic context in pathway analysis of GWAS data

In recent years, RNA-seq has become a very competitive alternative to microarrays. In RNA-seq experiments, the expected read count for a gene is proportional to its expression level multiplied by its transcript length. Even when two genes are expressed at the same level, differences in length will yield differing numbers of total reads. The characteristics of these RNA-seq experiments create a gene-level bias such that the proportion of significantly differentially expressed genes increases with the transcript length, whereas such bias is not present in microarray data. Gene-set analysis seeks to identify the gene sets that are enriched in the list of the identified significant genes. In the gene-set analysis of RNA-seq, the gene-level bias subsequently yields the gene-set-level bias that a gene set with genes of long length will be more likely to show up as enriched than will a gene set with genes of shorter length. Because gene expression is not related to its transcript length, any gene set containing long genes is not of biologically greater interest than gene sets with shorter genes. Accordingly the gene-set-level bias should be removed to accurately calculate the statistical significance of each gene-set enrichment in the RNA-seq.We present a new gene set analysis method of RNA-seq, called FDRseq, which can accurately calculate the statistical significance of a gene-set enrichment score by the grouped false-discovery rate. Numerical examples indicated that FDRseq is appropriate for controlling the transcript length bias in the gene-set analysis of RNA-seq data. To implement FDRseq, we developed the R program, which can be downloaded at no cost from http://home.mju.ac.kr/home/index.action?siteId=tyang.

Genome-wide association studies (GWAS) for epithelial ovarian cancer (EOC), the most lethal gynecologic malignancy, have identified novel susceptibility loci. GWAS for survival after EOC have had more limited success. The association of each single-nucleotide polymorphism (SNP) individually may not be well suited to detect small effects of multiple SNPs, such as those operating within the same biologic pathway. Gene set analysis (GSA) overcomes this limitation by assessing overall evidence for association of a phenotype with all measured variation in a set of genes. To determine gene sets associated with EOC overall survival, we conducted GSA using data from two large GWAS (N cases = 2,813, N deaths = 1,116), with a novel Principal Component-Gamma GSA method. Analysis was completed for all cases and then separately for high-grade serous histologic subtype. Analysis of the high-grade serous subjects resulted in 43 gene sets with P < 0.005 (1.7%); of these, 21 gene sets had P < 0.10 in both GWAS, including intracellular signaling pathway (P = 7.3 × 10(-5)) and macrolide binding (P = 6.2 × 10(-4)) gene sets. The top gene sets in analysis of all cases were meiotic mismatch repair (P = 6.3 × 10(-4)) and macrolide binding (P = 1.0 × 10(-3)). Of 18 gene sets with P < 0.005 (0.7%), eight had P < 0.10 in both GWAS. This research detected novel gene sets associated with EOC survival. Novel gene sets associated with EOC survival might lead to new insights and avenues for development of novel therapies for EOC and pharmacogenomic studies.

Attention-deficit/hyperactivity disorder (ADHD) and autism spectrum disorders (ASD) often co-occur. Both are highly heritable; however, it has been difficult to discover genetic risk variants. Glutamate and GABA are main excitatory and inhibitory neurotransmitters in the brain; their balance is essential for proper brain development and functioning. In this study we investigated the role of glutamate and GABA genetics in ADHD severity, autism symptom severity and inhibitory performance, based on gene set analysis, an approach to investigate multiple genetic variants simultaneously. Common variants within glutamatergic and GABAergic genes were investigated using the MAGMA software in an ADHD case-only sample (n=931), in which we assessed ASD symptoms and response inhibition on a Stop task. Gene set analysis for ADHD symptom severity, divided into inattention and hyperactivity/impulsivity symptoms, autism symptom severity and inhibition were performed using principal component regression analyses. Subsequently, gene-wide association analyses were performed. The glutamate gene set showed an association with severity of hyperactivity/impulsivity (P=0.009), which was robust to correcting for genome-wide association levels. The GABA gene set showed nominally significant association with inhibition (P=0.04), but this did not survive correction for multiple comparisons. None of single gene or single variant associations was significant on their own. By analyzing multiple genetic variants within candidate gene sets together, we were able to find genetic associations supporting the involvement of excitatory and inhibitory neurotransmitter systems in ADHD and ASD symptom severity in ADHD.

To date, genome-wide association studies have identified thousands of statistically-significant associations between genetic variants, and phenotypes related to a myriad of traits and diseases. A key goal for human-genetics research is to translate these associations into functional mechanisms. Popular gene-set analysis tools, like MAGMA, map variants to genes they might affect, and then integrate genome-wide association study data (that is, variant-level associations for a phenotype) to score genes for association with a phenotype. Gene scores are subsequently used in competitive gene-set analyses to identify biological processes that are enriched for phenotype association. By default, variants are mapped to genes in their proximity. However, many variants that affect phenotypes are thought to act at regulatory elements, which can be hundreds of kilobases away from their target genes. Thus, we explored the idea of augmenting a proximity-based mapping scheme with publicly-available datasets of regulatory interactions. We used MAGMA to analyze genome-wide association study data for ten different phenotypes, and evaluated the effects of augmentation by comparing numbers, and identities, of genes and gene sets detected as statistically significant between mappings. We detected several pitfalls and confounders of such “augmented analyses”, and introduced ways to control for them. Using these controls, we demonstrated that augmentation with datasets of regulatory interactions only occasionally strengthened the enrichment for phenotype association amongst (biologically-relevant) gene sets for different phenotypes. Still, in such cases, genes and regulatory elements responsible for the improvement could be pinpointed. For instance, using brain regulatory-interactions for augmentation, we were able to implicate two acetylcholine receptor subunits involved in post-synaptic chemical transmission, namely CHRNB2 and CHRNE, in schizophrenia. Collectively, our study presents a critical approach for integrating regulatory interactions into gene-set analyses for genome-wide association study data, by introducing various controls to distinguish genuine results from spurious discoveries.

Differential gene expression (DGE) studies often suffer from poor interpretability of their primary results, i.e., thousands of differentially expressed genes. This has led to the introduction of gene set analysis (GSA) methods that aim at identifying interpretable global effects by grouping genes into sets of common context, such as, molecular pathways, biological function or tissue localization. In practice, GSA often results in hundreds of differentially regulated gene sets. Similar to the genes they contain, gene sets are often regulated in a correlative fashion because they share many of their genes or they describe related processes. Using these kind of neighborhood information to construct networks of gene sets allows to identify highly connected sub-networks as well as poorly connected islands or singletons. We show here how topological information and other network features can be used to filter and prioritize gene sets in routine DGE studies. Community detection in combination with automatic labeling and the network representation of gene set clusters further constitute an appealing and intuitive visualization of GSA results. The RICHNET workflow described here does not require human intervention and can thus be conveniently incorporated in automated analysis pipelines.

RATIONALE: Although inhaled corticosteroids (ICS) are the most commonly prescribed medication for asthma, 35% of patients respond poorly to ICS. Therefore, higher ICS doses are required to control asthma symptoms, which can increase the risk of adrenal suppression caused by a reduction of endogenous glucocorticoid production. Recently, multi-omics provided a new opportunity to explore potential biomarkers that affect treatment-related outcomes. Here, we applied this approach to understand risk for developing adrenal suppression in asthmatic patients taking ICS. METHODS: Plasma metabolomics profiling and a gene set analysis using genome-wide genotype data were performed on 419 adult asthma patients of European ancestry from the Pharmacogenomics of Adrenal Suppression with Inhaled Corticosteroids (PhASIC) cohort, all of whom were on ICS at the time of sampling. We identified 810 metabolites and &amp;gt;3 million SNPs mapped to 26,458 genes that passed quality control filters. Using cortisol levels as a continuous outcome for adrenal status, multiple linear regression was performed to identify associations between metabolites/genes and outcome, followed by validation in a separate cohort of 315 patients. Using validated metabolites/genes, five machine learning models were developed and tested for potential to predict adrenal insufficiency: Regularized Logistic Regression; Support Vector Machine; Random Forest; Gradient Boosting Model; Multi-Layer Perceptron. For evaluation of predictive ability, patients were categorized into two groups based on plasma cortisol values as ‘adrenal sufficient’ (&amp;gt; 18 μg/dL) or ‘suspicion of adrenal insufficiency’ (≤ 18 μg/dL). Pathway and network analyses were applied to investigate altered pathways related to cortisol. RESULTS: Five significant (ENT 85%) and 60 nominally-significant (nominal-P &amp;lt; 0.05) metabolites were found in linear regression models. In gene-set analysis, 142 genes demonstrated a strong significant association with adrenal suppression. Using an independent cohort, we validated 19 metabolites and 76 genes at a nominal P &amp;lt; 0.05. For prediction of adrenal suppression, the Regularized Logistic Regression model combining these validated genes and metabolites achieved the highest AUC of 0.84. Joint pathway and network analyses revealed alterations in metabolism of steroids, long chain polyunsaturated fatty acids, and glycerophospholipids related to cortisol. CONCLUSIONS: Integration of genotyping data with global metabolomics identified significant features associated with continuous cortisol as a biomarker for adrenal suppression in asthma patients on ICS. A high accuracy of risk prediction for adrenal insufficiency could also be obtained using machine learning models that included genes and metabolites. Pathway analysis also identified fatty acids, steroids, and glycerophospholipid metabolism were changed in relation to adrenal suppression.